A Hybrid Tracking Method Based on Active Contour and Mean Shift Algorithm

Abstract

Active contour is a very accurate in target tracking and robust to variety of illumination, translation, rotation and large scale. Unfortunately, the problem that active contour method can not track target in real time because of huge computation is still not well resolved. On the contrast, mean shift is a very fast algorithm in target tracking but sensitive to the change of illumination, and can't get the description of target's shape. In our work, we use mean shift method to determine the translation motion of target and build a initial rough contour for active contour method. So the computation of active contour method in early curves evolution stage is greatly reduced and the target can be tracked accurately by minimizing the energy function in few number of interactive computation. Experimental results validate our method.